Papers
Topics
Authors
Recent
Gemini 2.5 Flash
Gemini 2.5 Flash
110 tokens/sec
GPT-4o
56 tokens/sec
Gemini 2.5 Pro Pro
44 tokens/sec
o3 Pro
6 tokens/sec
GPT-4.1 Pro
47 tokens/sec
DeepSeek R1 via Azure Pro
28 tokens/sec
2000 character limit reached

Joint Enhancement and Denoising Method via Sequential Decomposition (1804.08468v3)

Published 23 Apr 2018 in cs.CV

Abstract: Many low-light enhancement methods ignore intensive noise in original images. As a result, they often simultaneously enhance the noise as well. Furthermore, extra denoising procedures adopted by most methods ruin the details. In this paper, we introduce a joint low-light enhancement and denoising strategy, aimed at obtaining well-enhanced low-light images while getting rid of the inherent noise issue simultaneously. The proposed method performs Retinex model based decomposition in a successive sequence, which sequentially estimates a piece-wise smoothed illumination and a noise-suppressed reflectance. After getting the illumination and reflectance map, we adjust the illumination layer and generate our enhancement result. In this noise-suppressed sequential decomposition process we enforce the spatial smoothness on each component and skillfully make use of weight matrices to suppress the noise and improve the contrast. Results of extensive experiments demonstrate the effectiveness and practicability of our method. It performs well for a wide variety of images, and achieves better or comparable quality compared with the state-of-the-art methods.

User Edit Pencil Streamline Icon: https://streamlinehq.com
Authors (4)
  1. Xutong Ren (3 papers)
  2. Mading Li (7 papers)
  3. Wen-Huang Cheng (40 papers)
  4. Jiaying Liu (99 papers)
Citations (182)

Summary

Joint Low-Light Enhancement and Denoising via Sequential Decomposition

This paper introduces a novel approach for addressing the challenge of enhancing low-light images while simultaneously denoising them. The authors propose a joint enhancement and denoising framework using sequential decomposition, which aims to maintain image quality by effectively suppressing noise while enhancing the illumination of low-light images. Traditional low-light enhancement techniques often inadvertently amplify the inherent noise present in these images, necessitating subsequent denoising steps that can degrade image details. This paper addresses this challenge by integrating noise suppression directly into the Retinex-based decomposition process.

The authors identify the limitations of conventional methods that separate the enhancement and denoising processes, for instance, the classic Retinex model, which assumes that an image is a product of illumination and reflectance. By introducing a noise term into this model, the authors devise a method that simultaneously refines the illumination and reduces noise in the reflectance. Their proposed method operates in a two-step, sequential decomposition process, which first estimates an illumination map while suppressing noise and then refines the reflectance with minimal noise interference.

A primary innovation of this approach is the modification of the Retinex model to account for noise by incorporating a noise term. This approach explicitly considers noise as a factor during decomposition, which helps prevent enhancement of noise. The sequential estimation process leverages weighted matrices to smooth the illumination and suppress noise in the reflectance through spatial regularization. The result is a more robust illumination map and a clearer reflectance, which maintain a higher quality in the enhanced image without requiring additional denoising procedures.

The experimental results presented in the paper underscore the superiority of the proposed method over existing enhancement techniques. The method was benchmarked against several state-of-the-art enhancement approaches and was shown to produce enhanced images with superior detail retention and noise suppression. This robustness is attributed to the integrated treatment of noise within the sequential decomposition framework, avoiding the detail loss and over-enhancement issues seen with other methods.

The implications of this research are significant for practical applications where low-light imaging is necessary, such as photography in suboptimal light conditions, mobile device cameras, and video surveillance. By providing a method that balances enhancement and noise suppression, the work holds promise for integration into real-time imaging systems where preserving detail and minimizing noise are critical.

Theoretically, the introduction of noise as an integral part of the decomposition process may invite further exploration into joint modeling frameworks for other imaging tasks where blur, noise, or other artifacts coexist with the main signal. It opens avenues for enhancing imaging techniques in medical imaging, remote sensing, and multimedia processing.

Future research may explore the integration of this sequential decomposition method with advanced machine learning techniques to further refine the model's adaptability and improve its performance across varied imaging conditions. The exploration of real-time processing capabilities and extension to other types of image artifacts could also provide fertile ground for further academic inquiry. Overall, this work contributes a significant advancement in enhancing low-light images by effectively marrying enhancement and denoising within a unified framework.